Looking up from the bridge
INVESTMENT TRENDS

The freeway to an electric vehicle future

The electrification of vehicles and the ultimate transition to autonomous driving is happening at an accelerating pace. The transition to autonomous vehicles (AVs) has the potential to fundamentally change the transportation industry and the related investment landscape.

The electrification of vehicles and the ultimate transition to autonomous driving is happening at an accelerating pace. The transition to autonomous vehicles (AVs) not only changes the technologies involved but could also have significant societal impacts, potentially reducing car ownership and destroying brand value, changing the automotive ecosystem as we know it. This has the potential to fundamentally change the transportation industry and the related investment landscape.

Five levels of automation

We see safety and regulation, alongside consumer demand and decreasing costs, as being the key determinants of increased technology deployment in automotive. According to the National Highway Traffic Safety Administration in the United States, human error causes 94% of car accidents1. We anticipate that local regulation and specific driving conditions will cause the pace of autonomous driving adoption to vary by region. We don’t doubt that at some point fully autonomous vehicles will be ubiquitous, but the timing and route to such a transition is uncertain.

The Society of Automotive Engineers broadly defines five stages of autonomous driving, with Level 5 representing full autonomy. Level 1 is the most basic level and is mainstream in many car models today.

Five levels of automation

Five levels of automation

Source: UBS, Society of Automotive Engineers, as of October 2018. For illustrative purpose only. There is no guarantee that any forecast made will come to pass.

Gathering information – the eyes and ears

Firstly, an autonomous vehicle needs to gather information before it can action it, similar to how a human uses their eyes and ears whilst driving. From a technology perspective, this requires the use of sensors. The key technologies include radar, lidar and cameras each with their own strengths and weaknesses, such that we think all technologies will ultimately be required for full automation.

Higher levels of vehicle automation require large sensing and computer power, creating a significant opportunity for companies supplying these technologies.

Radar, Camera and Lidar for different levels of automation

Source: Infineon, as of October 2018.

Processing information – the brain

Sensor fusion is a process which encompasses the information from all three technologies mentioned above. In addition to machine learning, sensor fusion enables the vehicle to process the information that the sensors gather and make informed decisions, similar to how a human brain operates. From a technology standpoint, processing the data collected requires both advanced software and significant computing capabilities:

  • Software: Artificial intelligence and deep learning, a sophisticated form of artificial intelligence, which closely replicates the decision-making process of a human brain. The process of deep learning analyses large data sets using complex algorithms. In practice, this means simultaneously processing images that the aforementioned sensors detect and determining whether the vehicle needs to take action, such as slowing down or steering.
  • Mapping: There are currently two different approaches to achieving Level 4 & 5 autonomy. The first is a mapbased approach, where high-definition (HD) maps enable the car to navigate itself by comparing the map with real time data, most likely collected by lidar. The second approach is camera-centric and more evolutionary in nature, which relies on camera and radar to make informed decisions and enables vehicles to gradually automate with incrementally more functions becoming automated.
  • Hardware:Analysing large data sets extremely quickly is very compute intensive and required a breakthrough in the semiconductor industry. This was achieved by “massively parallel” semiconductor architectures which in practice means the data set is broken down into more manageable sub-sets which are analysed and executed at the same time. This represented a leap in technology from the incumbent central processing unit (CPU), which proved too slow to keep up with the development of the needs currently.

Where are the investment opportunities?

Green Cargo truck in container port

Although the path of the transition and the timing is uncertain, many of the technologies exist today but will become more complex and more widely used.

Certain semiconductor companies are set to outgrow the automotive market for a multi-year period as content per vehicle rises. Near term we see this being driven by sensors, which the vehicles will rely on to gather data. Longer term, as software becomes increasingly complex it is likely that less sensors will be required and software will capture increasing value in the chain. Many of these technologies are in their infancy today, creating an opportunity for active investor to identify companies which are in the best position to benefit longer-term.